Skip to content
Back to skills

Ltm Init

ASecurity

Initialize the claude-ltm long-term memory system in the current project

  • 4 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added May 26, 2026
documentationdatabase

Security analysis

A100/100

Scanned May 27, 2026

npx -y skills add LaserPhaser/claude-ltm --skill ltm-init --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Ltm Init?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Ltm Init
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/laserphaser-ltm-init/badge)](https://www.skillsdirectory.com/skills/laserphaser-ltm-init)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: ltm:init
description: Initialize the claude-ltm long-term memory system in the current project
---

# Initialize Long-Term Memory

Set up the claude-ltm memory system in this project.

## What to do

1. Check if `.memory/DECISIONS.md` already exists. If it does, inform the user
   that LTM is already initialized and ask if they want to re-initialize
   (which will update CLAUDE.md instructions but NOT overwrite
   existing decisions).

2. Create the following structure:
   ```
   .memory/
   ├── DECISIONS.md    (empty memory with Status + Changelog scaffolding)
   └── details/        (empty directory for future detail files)
   ```

3. Create `.memory/DECISIONS.md` with this content:
   ```markdown
   # Project Memory

   > Auto-maintained by claude-ltm. Last updated: YYYY-MM-DD

   ## Status
   - **Focus**:
   - **Blockers**: None
   - **Open**:

   ## Changelog
   ```

4. Add the LTM instructions section to the project's `CLAUDE.md`. If CLAUDE.md
   doesn't exist, create it. If it exists, append the LTM section at the end.
   Do NOT duplicate if `<!-- BEGIN claude-ltm -->` already exists.

   The instructions to add (wrapped in markers for clean uninstall):
   ```
   <!-- BEGIN claude-ltm -->
   # Long-Term Memory (claude-ltm)

   You have a project decision memory at `.memory/DECISIONS.md`. This file is the
   team's persistent record of significant decisions made during development.

   ## At session start
   - Read `.memory/DECISIONS.md` to understand prior decisions and current status.

   ## During work
   - When you make or discover a significant decision (architecture, tech stack,
     changed approach, business logic, resolved debate), update `.memory/DECISIONS.md`
     immediately.
   - "Significant" = something a teammate joining next week would need to know.
     Routine code changes are NOT decisions. Choosing a database IS.
   - Each entry gets a ### heading with [date], up to ~10 lines of context.
   - If a decision is truly complex (>10 lines), create a detail file in
     `.memory/details/YYYY-MM-DD-<topic>.md` and link to it.
   - When a prior decision is superseded, update its entry in-place. Note what
     changed and why. Update the Changelog section.
   - Check for duplicates before adding — update existing entries, don't repeat.

   ## Before finishing your session
   - Do a compact pass: review this session's work and check if any decisions
     were made but not yet recorded in `.memory/DECISIONS.md`.
   - Update the Status section (current focus, blockers, open questions).
   - Skip if nothing significant was decided.
   <!-- END claude-ltm -->
   ```

5. Confirm to the user what was created and that LTM is now active.

## After initialization

Read `.memory/DECISIONS.md` to confirm it's properly set up. From now on,
follow the LTM instructions in CLAUDE.md for maintaining the decision memory.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.